Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallIf ONNX Runtime Web is asked to use multiple WebAssembly threads on a page that is not cross-origin isolated, the runtime can continue with one thread instead. In a benchmark reported by hao jia on September 29, 2026, the four-thread request without isolation took a median 6,318 ms, close to the 6,313 ms isolated single-thread result; four threads with isolation took 2,133 ms. Those are results for one specific setup, not a general speedup guarantee.
Why does `numThreads = 4` still run on one thread?
`numThreads` is a request, not a way to make an unsupported browser environment provide WebAssembly multithreading. ONNX Runtime Web’s documented requirements are browser support for WebAssembly multithreading and an enabled `crossOriginIsolated` mode. When the runtime detects that multithreading is unavailable despite a request for more than one thread, its current upstream implementation warns and resets the effective count to one before initialization. The source warning is: “WebAssembly multi-threading is not supported in the current environment. Falling back to single-threading.”
The flag documentation also clarifies that the count includes the main thread: a value of one disables multithreading. A value of zero delegates thread selection to the environment; the documented browser choice is half of `navigator.hardwareConcurrency` or four, whichever is smaller. Set the flag before creating the inference session. The implementation and warning are from the mutable upstream `main` branch, so behavior and wording may change between releases.
Sources: ONNX Runtime WebAssembly flags documentation and upstream WASM factory source.
#1 Best Overall
- The world’s fastest gaming processor, built on AMD ‘Zen5’ technology and Next Gen 3D V-Cache.
- 8 cores and 16 threads, delivering +~16% IPC uplift and great power efficiency
- 96MB L3 cache with better thermal performance vs. previous gen and allowing higher clock speeds, up to 5.2GHz
- Drop-in ready for proven Socket AM5 infrastructure
- Cooler not included
What do COOP and COEP change?
Cross-origin isolation is a browser page state, not an ONNX Runtime setting. A cited reproduction enabled it by serving these response headers on the page:
Cross-Origin-Opener-Policy: same-origin
Cross-Origin-Embedder-Policy: require-corpIn that reproduction, `crossOriginIsolated` became true when both headers were sent and false when neither was sent. The exact policy needed in production depends on the page and its cross-origin resources, so test the actual deployed page rather than assuming the development server’s state carries over. Check `globalThis.crossOriginIsolated` in the browser console, and inspect response headers at the layer that serves the page, such as the web server or CDN.
Rank #2
- AMD Ryzen 9 9950X3D Gaming and Content Creation Processor
- Max. Boost Clock : Up to 5.7 GHz; Base Clock: 4.3 GHz
- Form Factor: Desktops , Boxed Processor
- Architecture: Zen 5; Former Codename: Granite Ridge AM5
Source: hao jia’s benchmark and reproduction.
How large was the reported performance difference?
In the author-reported benchmark posted September 29, 2026, the setup was `onnxruntime-web` 1.27.0, an ISNet INT8 model, a 1024 × 1024 input, an Apple M4 with 16 GB of memory, and an open-source Chromium 149 build. The measurements below are median steady-state inference times reported for that setup.
| Page state and setting | Reported median |
|---|---|
| Cross-origin isolated; four threads | 2,133 ms |
| Not cross-origin isolated; four threads requested, runtime falls back to one | 6,318 ms |
| Cross-origin isolated; one thread | 6,313 ms |
The non-isolated run completed without an exception, printed two warnings, and was close to the isolated one-thread result. This supports the diagnosis that a successful inference call does not prove the requested thread count took effect. It does not establish how often this occurs across deployments, or a universal speedup ratio: the figures are one developer’s benchmark, not an independently replicated study.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteRank #3
- Can deliver fast 100 plus FPS performance in the world's most popular games, discrete graphics card required
- 6 Cores and 12 processing threads, bundled with the AMD Wraith Stealth cooler
- 4.2 GHz Max Boost, unlocked for overclocking, 19 MB cache, DDR4-3200 support
- For the advanced Socket AM4 platform
Source: benchmark details and results.
How can you make degraded mode visible?
Choose an explicit count from the page’s isolation state before creating the session, and send an application-owned diagnostic to your logging or telemetry system when isolation is unavailable. That gives you a visible signal beyond a browser console warning and makes the intended configuration inspectable.
ort.env.wasm.numThreads = globalThis.crossOriginIsolated ? 4 : 1;
if (!globalThis.crossOriginIsolated) {
// Replace with your application's logging or telemetry call.
console.warn("Cross-origin isolation is unavailable; using one WASM thread.");
}
const session = await ort.InferenceSession.create(modelUrl);The four-thread choice above is an example matching the cited benchmark, not a universal optimum. The report observed `{ isolated: true, threads: 4 }` on its isolated test server and `{ isolated: false, threads: 1 }` on the other. In your application, verify the actual deployment state and monitor the diagnostic; do not infer effective threading solely from the configured value.
Rank #4
- Pure gaming performance with smooth 100+ FPS in the world's most popular games
- 6 Cores and 12 processing threads, based on AMD "Zen 5" architecture
- 5.4 GHz Max Boost, unlocked for overclocking, 38 MB cache, DDR5-5600 support
- For the state-of-the-art Socket AM5 platform, can support PCIe 5.0 on select motherboards
- Cooler not included
Source: tested configuration and mitigation.
How should you verify the performance impact?
- Confirm `globalThis.crossOriginIsolated` on the page where inference runs.
- Record the requested thread count and the isolation state alongside application telemetry.
- Compare the same model, input dimensions, execution provider, browser, device, and deployment conditions.
- Measure warm or steady-state inference consistently; do not compare a cold start with a warmed run.
- Test the configuration users actually receive, including the production response headers and relevant cross-origin resources.
There is no source-backed universal best thread count or general speedup figure. The right choice depends on the environment and workload being measured.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Does this affect WebGPU or proxy workers?
The reported fallback is about WebAssembly threading. ONNX Runtime documents WASM as its default CPU provider and WebGPU as a distinct execution provider. In the same author-reported test, WebGPU medians were 355 ms without isolation and 359 ms with isolation. That limited comparison does not prove WebGPU is always unaffected by isolation; test the provider and deployment you use.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Best Value
- Processor provides dependable and fast execution of tasks with maximum efficiency.Graphics Frequency : 2200 MHZ.Number of CPU Cores : 8. Maximum Operating Temperature (Tjmax) : 89°C.
- Ryzen 7 product line processor for better usability and increased efficiency
- 5 nm process technology for reliable performance with maximum productivity
- Octa-core (8 Core) processor core allows multitasking with great reliability and fast processing speed
- 8 MB L2 plus 96 MB L3 cache memory provides excellent hit rate in short access time enabling improved system performance
A proxy worker is a separate option aimed at keeping heavy work from blocking the main thread, which can improve UI responsiveness. ONNX Runtime’s performance guidance does not describe it as a way to improve model performance, and it does not fix the missing cross-origin isolation needed for WASM multithreading. The flag documentation also notes proxy-worker limitations with WebGPU and pages restricted by Content Security Policy.
Sources: WebAssembly flags documentation, performance and threading guidance, environment flags and session options, and benchmark report.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
